Image processing based on matlab full variation image processing matlab source code 457

function compareimages(A,ATitle,B,BTitle)
%COMPAREIMAGES   Displays two images side by side with linked axes
%   COMPAREIMAGES(A,B) displays images A and B, where A and B are either
%   grayscale orRGB color images with values in [0,1]for convenient panning and zooming.
%
%   COMPAREIMAGES(A,'A title',B,'B title') specifies titles above the
%   images.
%
%   See also linkaxes.

% Pascal Getreuer 2009

if nargin = =2
    B = ATitle;
    ATitle = "'; BTitle = ''; elseif nargin < 4 error('Must have 2 or 4 input arguments. ');
end

ax(1) = subplot(1.2.1);
hold off

if ndims(A) = =2
    image(A*255);
    colormap(gray(256));
elseif ndims(A) = =3
    image(min(max(A,0),1));
end

set(gca,'Units'.'Normalized'.'Position'[0.0.1.0.5.0.8]);
axis image
axis off
title(ATitle);
zoom;

ax(2) = subplot(1.2.2);
hold off

if ndims(B) = =2
    image(B*255);
    colormap(gray(256));
elseif ndims(B) = =3
    image(min(max(B,0),1));
end

set(gca,'Units'.'Normalized'.'Position'[0.5.0.1.0.5.0.8]);
axis image
axis off
title(BTitle);
%%% Demo of image deconvolution %%%

BlurRadius = 3;
NoiseLevel = 0.005; 
lambda = 4e3;

uexact = double(imread('einstein.png')) /255;

% Construct the blur filter
[x,y] = meshgrid(1:size(uexact,2),1:size(uexact,1));
psf = double((x-size(uexact,2) /2). ^2. + (y-size(uexact,1) /2). ^2 <= BlurRadius^2);
psf = psf/sum(psf(:));

% Simulate a noisy andblurry image f = real(ifft2(fft2(uexact).*fft2(fftshift(psf)))); f = f + randn(size(uexact))*NoiseLevel; % Deblur u = tvdeconv(f,lambda,psf); unction u = tvdenoise(f,lambda,varargin) %TVDENOISE Total variation image denoising. % u = TVDENOISE(f,lambda,model) denoises grayscale, color,or arbitrary
%   multichannel image f using total variation regularization.  Parameter
%   lambda controls the strength of the noise reduction: smaller lambda
%   implies stronger denoising.
%
%   The model parameter specifies the noise model (case insensitive):
%     'Gaussian' or 'L2'  - (default) The degradation model for additive
%                           white Gaussian noise (AWGN),
%                             f = (exact) + (Gaussian noise).
%     'Laplacian' or 'L1' - The degradation model assumes impulsive noise, 
%                           for example, salt & pepper noise.
%     'Poisson'- Each pixel is an independent Poisson random % variable with mean equal to the exact value. % % TVDENOISE(... ,Tol,MaxIter) specify the stopping toleranceand the
%   maximum number of iterations.
%
%   See also tvdeconv, tvinpaint, and tvrestore.
Copy the code

Version: 2014 a